Name: Brainstorm Health 2024: Defining AI Standards for the Next Era of Health
Uploaded: 2024-05-21
Duration: 18 min 54 s
Description: Aneesh Chopra, President, CareJourney Lloyd Minor, Carl and Elizabeth Naumann Dean, Stanford University School of Medicine; Vice President, Medical Affairs, Stanford University Moderator: Deena Shakir, Lux Capital

Aneesh Chopra, President, CareJourney Lloyd Minor, Carl and Elizabeth Naumann Dean, Stanford University School of Medicine; Vice President, Medical Affairs, Stanford University Moderator: Deena Shakir, Lux Capital

Brainstorm Health 2024: Defining AI Standards for the Next Era of Health

Conferences

Aneesh Chopra, CareJourney, Lloyd Minor, Carl and Elizabeth Naumann Dean, Stanford University School of Medicine; Vice President, Medical Affairs, Stanford University, Deena Shakir, Lux Capital

Hello, everyone. Delighted to have this conversation with two such wonderful panelists, Anesh and I go way back to our days uh in the early days of the Obama administration and doctor Miner needs no introduction. So today we're here to talk about the acronym on everyone's minds at whatever conference you're at these days. But in particular, at the intersection of A I and health, uh I want to start with a framing question. And so let's start with you Anish, what is A I in the context of health care? Well, to me, there's basically a policy construct for A I and health, which is to realize the promise that we made, we made the investments in electronic health records that is we promised a productivity revolution that would make the system better. We invested in the technology, we've not yet gotten the productivity gains. And so A I is essentially the maybe the next chapter to finish the story of what we initially invested in, which is, it's the engine that will drive productivity and health care. That's sort of one aspect of it. The more pragmatic aspect of it is it's a prediction of a next event. And one can look at that by looking at historical data and then sort of drawing a line using math to get to the next event or with these new exciting large language models to basically see what happens when you throw all of the information in the internet into a system to help it figure out what goes next. And it looks like a reasoning engine, which is what has cut the excitement for all of us today. You know, if you ask any doctor and there are many of you here in the audience about their experience with, with health care. It's often not a positive one. I come from a family of doctors. And I distinctly remember the day my father had to switch his E hr and how frustrated he was and yet the promise of technology has the potential to revolutionize productivity of the health care workforce. When I think about A I and health care, I remember the early announcement of doximity and how this very same sentence where doximity was announcing an integration with chat GPT included the word fax machine. And I thought that is the most health care example that I have ever heard in terms of technology and A I dr minor in the context of the work that you are leading at Stanford and in particular the conference that you held last week. How are you thinking about A I in health education specifically? Sure. Thank you. And I think there's an enormous amount of green space here, right, in terms of the impact that A I is going to have throughout the continuum of medical education care delivery, discovery based science. And also importantly access to care. Uh We talked a lot about efficiency of of health care, talked about improving the quality of the health care delivery. But also, you know, A I has the potential to dramatically increase the access to specialists, the access to the type of specialty care that oftentimes is only granted today, only given today in centers in the educational sphere. You know, we're educating physicians and scientists for the next 3040 years of their productive lifetime, medical education, even scientific education, biomedical science used to be heavily based upon memorization and understanding of a core basis of facts that's been diminished. But I think it's going to be diminishing even more as generative A I becomes multimodal, generative A I and incorporates more than language, incorporates images, incorporates data from wearables and other information that can be assimilated in ways that today we depend upon humans to do it. Whether or not those humans are physicians or scientists or a mixture of the two. That's really an exciting space for us. I think education has received overall less attention uh today than some of the other applications we've talked about. But the potential is just as enormous. One area that I wanted to get both of your thoughts on in particular is this question of bias in A I which is uh which is the topic I know that is top of mind, certainly at Stanford and, and for the work that you're doing and niche, we talk about garbage and garbage out. If A I is mirroring human bias, how can we mitigate that impact as A I replicates that scale? So we spend a lot of time on this question because we deal with regulating inputs. That's one question, how do you regulate models? But we also think about managing outputs and so acknowledging biases, but then building processes in place to address them may actually be the key. So if you look at the home page of the White House website on algorithmic discrimination, it's the poster child of the Optim algorithm that allocated care management resources to based on spend which you know, incorporated the bias of where and how people access health care. But if you focus on the outcome, which is who has the highest and best need and how do we allocate limited resources to address it? One can build outcomes programs to deal with this issue. Even if the model itself may not have the right information, by the way, that is a part and parcel of why over 40 systems and health plans committed at the White House in December that they would take on a more of a self regulatory process to focus on the outcomes and the use of A I as their, as their obligation rather than just worrying about what's in the, er, or what's in the software. I think outcomes focus is the key with the focus on outcomes is value based care. A better use case for A I. Well, we're going to get that in a minute. But absolutely, if you said, what's the productivity function of the health care system at the hospital? It might be nurse productivity to get staffing ratios down. If it's a health plan, it might be medical loss ratio more broadly. In the case of like, what do you want your mom and dad to get for the best possible care? You want them to live longer and healthier and you want them to be on track with all their preventive services. A value based care organization's key performance indicators, productivity improvement on those aligns best to societal goals. In my view as a, as a VC, it's my job to be an optimist. But I think we need to have a conversation about what could go wrong in this case. And as we were chatting earlier, you know, I think back to the early 20 tens and the promise of social media and technology as a democratizing force and a force for good years later, we look back and realize there were unintended consequences and it's not always going to go the way that we anticipate. As we think about the rapid deployment of A I and the rapid innovation in this space. How can we mitigate future unintended consequences, Dr Mor? Well, I think one important topic is how do we think about the responsible role of regulation and compliance? Um And I've been very impressed, not surprised but very impressed about how federal agencies, the office of National coordinator of the FDA are really leaning in and wanting to understand more about how systems today are deploying A I. Um how A I models are validated, what, how that validation may need may need to be incorporated into a national regulatory framework. But to the analogy with social media and what went wrong in a sense and what can be prevented? Um I think with health care from the get go, we will have more engagement appropriately from regulatory agencies than we've had in social media because of some of the legal constructs protecting companies from liability. On the way the platforms are used, certainly in health care, there will be decisions made about how liability is apportioned, but there's no question that systems doctors are responsible for the care that's being delivered, whether or not that care is being augmented and hopefully improved A I or not. So that's one I think key difference is we'll have more engagement appropriately from federal regulatory agencies to help drive this dialogue and make sure that we're focusing on consequences that could result from models are only going to be as good in making predictions as the data that they're fed, as you said, A I is in a very real sense, a mirror as to what it's learning and if what it's learning is from a biased data set, then its output is going to be biased. Um So understanding those processes and then having appropriate regulatory oversight, I think it's going to be really important moving forward. Maybe Dean, if I could just build on, on the dean's thoughtful comments, I think the engagement with regulators is a very positive, but it's a takes two to tango situation. We also need early adopters to collaborate in a more of a think of it, more like instead of a two sides to the negotiating table where I'm representing the folks who are using it and you're the regulator, we sort of have to be more at a table like we see in this room where we're, we're all rolling up our sleeves to learn together. So Harvard Business School and uh uh Boston Consulting Group did a study and they found that actually the smartest analysts were dumbed down, not enhanced by A I because they defaulted to the answer and sometimes the answer wasn't as good as what the best performers could do. So that was a head scratcher. What does that mean? Well, the regulators are not going to get to that level of minutia in what the regulation should say, be mindful of dumbing down. So you want stakeholders to set norms So University of California at San Diego said when we do the inbox draft notes and I think Stanford does this too to send messages to patients. There's no button that says submit. So you get the auto generated, there's no like send without it. The words are start from scratch or edit from where the A I left off. That kind of default setting is a norm or a behavior that is designed to mitigate against a learning that sometimes these things may have a negative effect. So as we learn together and we share best practices, there may be more coming out of the normative behavior of what should be the way we do this with the regulator saying thank you. Keep going. Relative to say the idea that we're waiting for a rule that may take a year and a half to get through the process to actually close whatever the perceived gaps are in just a moment, I'll turn to the audience for questions. So raise your hand and somebody will come to you with a mic. But in the meantime, I want to hear more specific examples about what you're excited about because A I can mean so many things and I think to, to, to an individual, sometimes A I and health care means an A I chatbot that's taking care of you. And we know that there's so much more to it than that. So can you share an example Dr Minor of an A I use case that you're excited about. Sure. And let me first follow up on a really important point that Anish made and that is a really important role for convening right now and for sharing ideas. Last week, we hosted at Sanford four days of conferences related to A I is deployment in health care and biomedicine beginning with a conference related to our initiative, raise health responsible A I for safe and equitable health. So if you ask me what I'm excited about, I'm excited about the way people are coming together, bringing their problems, their ideas and collectively reasoning how, you know, we can come together to solve those problems. This isn't going to be, you know, one massive deployment. It's going to be a lot of incremental deployments. It's going to be a lot of incremental advances but sharing ideas, what's working, what's not working, having methods for validating A I models before they're deployed as the coalition for Health A I or C which we convened the first meeting of that group last week at Stanford as well. That's going to be the way in which we determine the best applications and make sure that A I is being deployed responsibly and achieving the benefits that Anish and you and so many others have talked about that we should be attaining from A I absolutely question from the audience. Yes, back there. So we've heard a lot about A I in terms of productivity and improving the patient doctor relationship. Um My name is Rami Baloi. I'm the CEO of an no, but also a doctor who's been using A I. My question is, is what is the chances that A I by making certain administrative and third party services within healthcare so cheap that actually cements bad habits and uh stifles innovation. And to give you an example, um probably one of the largest growth areas in A I and health care that's not been talked about is billing coding, prior authorizations, prior authorization denials. And I would say this is even pre chat GP T and I will tell you the number of prior authorization denials that I've gotten that are A I generated and clearly missed the mark. This year has expanded exponentially. And so when those administrative burdens become so cheap, does it then decrease the risk taking of organizations to adopt a totally different model to go away from a framework that clearly has sort of stifled innovation in the past? If you don't mind, I might just take a stab because I think this gets to the nature of the question of the micro versus the macro policy objective and the goal. So each institution is going to have some use case in their administrative performance where they can use A I to be a little bit better. So that chart review to find that, you know, denial generating reason that's going to happen at scale, right? Now it's pecked because it's an expensive thing for plans to pay for that's gonna happen everywhere. Conversely, you're going to appeal every denial. So we can assume that in a micro economy, this is not the highest and best use of this incredible technology more broadly. What's the number one preventable death in this country? It's still hypertension. It's still uncontrolled hypertension. We launched a Million Hearts campaign in the beginning of Obama. If you remember that and we after all this, it investment, it's still bad. Ok. So what do we do to help people manage their hypertension? It's not, doesn't need, you know, GLPs or anything, you know, this is like simple dollar a day, aspirin sort of thing. So if you look at the productivity function, we want, that's what the Affordable Care Act is helping us to build towards. It was to look at these broader sort of lower quality outcomes that we're seeing in our system and then trying to incentivize organizations to own that as their productivity function. So let me take responsibility for a community to make sure that folks that have uncontrolled hypertension are better managed and they avoid the hospital. So I'm looking for what I refer to as a health information, fiduciary who's assigned with all the information necessary to identify where someone's falling off the tracks and then either directly nudging the individual patient or through a care manager or a physician practice brings them in to manage their care over time. That's the excitement that I see that will have the greatest societal good. We need to have economic model that rewards organizations for putting that service into play. But this is the big moment, institutional micro suboptimal administrative efficiencies A I versus patient centered macro. What do we want the world to be fewer deaths due to uncontrolled hypertension and so forth? And that to me is where I'm going to bet, I'm going to bet we're going to figure this out faster than just tweaking this. It's time for one more question and I see one over here. Hi, Eric J Daza here. So I'm a statistician. I'm bullish on A I, right. Statistics is the engine of the A I Mobile. Um But I'm also a doctor of Public Health and I worry about things like equity in this case, retraining and job training. So you've talked about job training using these A I tools. Can you talk a little bit about um, oh, I have this new A I tool. I don't want it to replace me or like a lot of my skills. How can I get retrained on that so I can still make a living? Thank you. You sure. Well, I think in every advance in technology, we've seen changes in the workforce. Um I, I think it follows on a very important point that Anish made. Um if deployed correctly, I think there will be that type of retraining that goes on in the workforce and will be able to get back in the delivery of patient care both in an individual basis. And at a larger public on a larger public health basis, will be able to get back to the type of interpersonal interactions that historically have been the bedrock strength of our health care delivery system. And that unfortunately, both through the way the system has evolved and also through the implications of current technology, we've been oftentimes separated from our patients. So will there be changes in the workforce? I think undoubtedly, I think there's no way to slow that down really other than to do it responsibly and appropriately. But if done appropriately, I think the workforce can be more effectively, far more effectively deployed in a more humanistic way, in a more rewarding way for workers than exists today where there's so much work that's just devoted to the administrative functions of the system, not directly to improving the lives of patients and populations. With just over a minute left, I'm going to do a fire round. I'm gonna, I'm gonna ask each of you to just say bearish or bullish on this particular use case with A I. So let's let's keep it to one word, bearish or bullish A I chatbot, bullish. Is that one word bullish prior, Bearish imaging, bullish, bullish, definitely voice to script bullish, bullish documentation related bullish also bullish. E hr innovation needed, needed un bullish, needed that bullish mental health, bullish. But man, that's, it's a new model. Likewise, bullish nursing has to be bullish. I hope. I mean, I'm, the protests are scaring me a little bit. So that's why I have a question mark. The uh the, the nurses are protesting the use of A I so anxiety. But I think we overcome this. I hope we'll learn. That's more than one word, pediatrics, I suppose. But bullish, I don't know. The use case specific is hard. Yeah, that's true. But bullish if it, if it helps to, you know, we've assumed all too commonly that kids are just small adults and that's not true. So if it helps us actually tailor health care appropriately for Children, then definitely bullish. Excellent capital b, great, great way to end. Thank you all so much. Fantastic panel.